[luckyq]

Series map (Phase 1 harness, one PR per step; each is based on the previous one)

PR	Step	Test cases
#137	Harness core: stack, client, telemetry, reference engine, CI unit tests; phase1 job in vllm-e2e.yml on the two-GPU runner (TC-01; #138/#139 extend --cases)	TC-01 integration generation correctness
#138	Local DRAM flow and hinted peer transfer, hybrid geometry flags	TC-02, TC-03
#139	Guard protection, takeover, replacement	TC-04
#140	Streaming client, performance arms, Guard recovery timing	TC-05, TC-06
#141	Sidecar kvcr_service, 2P+1D, two nodes, k8s / Slurm / Docker runners; phase1-2p1d CI job (2P+1D on the two-GPU runner, prefill workers share GPU 0)	all cases on those topologies
#142	docs/ci.md: what the two CI lanes verify, how to run them elsewhere, how to read evidence	—
#143	Test plan, run history, bare-vLLM probes	—
#180	--engine sglang: dynamo.sglang + HiCache KVCR storage backend (quick-start config), SGLang telemetry mapping, deterministic mode; no Guard/disagg	TC-01..03, TC-05 on SGLang

Replaces #103. All seven are drafts until the per-PR GPU runs on the current main image are done.

Problem

• Part 7 (last) of the stacked Phase 1 series. The harness, topologies and CI job are in; the test plan that maps requirements to cases, the run history with findings, and the bare-vLLM probes behind the hybrid-model analysis have no home in the repo.

What this adds

• docs/test-plan-phase1.md: case mapping (TC-01..06 → requirements and checks), topology and configuration, sequence, evidence layout, how to run (container, Docker wrapper, dlcluster, K8s), known limits, execution history on KVCR v0.1.1 (dense and hybrid, 2P+1D single and two nodes), findings (First peer transfer after Guard promotion or worker replacement is refused (missing target agent metadata) #51, MultiConnector top-level config, decode keeps a dead prefill's KV mapped, hybrid SSM checkpoint over-transfer), the review-driven verdict tightening, and candidate cases for the hint dispatcher (Add a KV hint dispatcher for multi-hint envelopes #118).
• scripts/probes/: three bare vllm serve probes that established the engine facts the hybrid verdicts rely on (one SSM checkpoint per request; junction checkpoint on the first fork; offload events per KV cache group before/after [KV Offload] Emit self-describing events for recurrent state groups vllm-project/vllm#62613).

Validation

• Docs only; ruff on the probes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.